Group items tagged

Generally these papers are just pumped-up versions of the five-paragraph essay, with filler added. Thesis-driven, argumentative, like the newspaper editorials the genre is based on, this "researched writing" promises to solve big questions with little effort: "Reproductive rights resolved in five pages!"

Actual writing related to research is modest, qualified, and hesitant

our actual model involves elaborately respectful conversation, demonstrating sensitivity to the most nuanced claims of previous researchers

Academic, legal, medical, and business writing has easily understandable conventions. We responsibly survey the existing literature, formally or informally creating an annotated bibliography. We write a review of the literature, identifying a "blank" spot ignored by other scholars, or a "bright" spot where we see conflicting evidence. We describe the nature of our research in terms of a contribution to the blank or bright spot in that conversation. We conclude by pointing to further questions.

Millions of pieces of research writing that aren’t essays usefully circulate in the profession through any number of sharing technologies, including presentations and posters; grant and experiment proposals; curated, arranged, translated, or visualized data; knowledgeable dialogue in online media with working professionals; independent journalism, arts reviews, and Wikipedia entries; documentary pitches, scripts and storyboards; and informative websites.

real researchers don’t write a word unless they have something to contribute. We should teach our students to do the same

LLMs usually cannot do a good job of explaining how a particular passage from a longer text illuminates the whole of that longer text. Moreover, ChatGPT’s outputs on comparison and contrast are often superficial. Typically the system breaks down a task of logical comparison into bite-size pieces, conveys shallow information about each of those pieces, and then formulaically “compares” and “contrasts” in a noticeably superficial or repetitive way. 

In-class writing, whether digital or handwritten, may have downsides for students with anxiety and disabilities

ChatGPT can produce outputs that take the form of  “brainstorms,” outlines, and drafts. It can also provide commentary in the style of peer review or self-analysis. Nonetheless, students would need to coordinate multiple submissions of automated work in order to complete this type of assignment with a text generator.  

No one should present auto-generated writing as their own on the expectation that this deception is undiscoverable. 

LLMs often mimic the harmful prejudices, misconceptions, and biases found in data scraped from the internet

Show students examples of inaccuracy, bias, logical, and stylistic problems in automated outputs. We can build students’ cognitive abilities by modeling and encouraging this kind of critique. Given that social media and the internet are full of bogus accounts using synthetic text, alerting students to the intrinsic problems of such writing could be beneficial. (See the “ChatGPT/LLM Errors Tracker,” maintained by Gary Marcus and Ernest Davis.)

Since ChatGPT is good at grammar and syntax but suffers from formulaic, derivative, or inaccurate content, it seems like a poor foundation for building students’ skills and may circumvent their independent thinking.

Good journalism on language models is surprisingly hard to find since the technology is so new and the hype is ubiquitous. Here are a few reliable short pieces.     “ChatGPT Advice Academics Can Use Now” edited by Susan Dagostino, Inside Higher Ed, January 12, 2023  “University students recruit AI to write essays for them. Now what?” by Katyanna Quach, The Register, December 27, 2022  “How to spot AI-generated text” by Melissa Heikkilä, MIT Technology Review, December 19, 2022  The Road to AI We Can Trust, Substack by Gary Marcus, a cognitive scientist and AI researcher who writes frequently and lucidly about the topic. See also Gary Marcus and Ernest Davis, “GPT-3, Bloviator: OpenAI’s Language Generator Has No Idea What It’s Talking About” (2020).

“On the Dangers of Stochastic Parrots” by Emily M. Bender, Timnit Gebru, et al, FAccT ’21: Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, March 2021. Association for Computing Machinery, doi: 10.1145/3442188. A blog post summarizing and discussing the above essay derived from a Critical AI @ Rutgers workshop on the essay: summarizes key arguments, reprises discussion, and includes links to video-recorded presentations by digital humanist Katherine Bode (ANU) and computer scientist and NLP researcher Matthew Stone (Rutgers).

at least users can see exactly how the sausage is made. Contrast this to Google or Bing, where the search algorithm that produces results is a closely guarded secret.

The word that Zucca used a number of times in our chat was "authority," and OUP is betting that individual and institutional users will value the authority enough that they'll be willing to pay for access to the service

This paywall is the only feature of OBO that seems truly unfortunate, given that the competition (search and Wikipedia) is free. High school kids and motivated amateurs will be left slumming it with whatever they can get from the public Internet, and OBO's potential reach and impact will be severely limite

Jarvis never broaches such subtleties. His is a simple world:

why not consider the possibility that the incumbents may be using the same tools, Jarvis’s revered technologies, to tell us what to think, and far more effectively than before? Internet shelf space may be infinite, but human attention is not. Cheap self-publishing marginally improves one’s chances of being heard, but nothing about this new decentralized public sphere suggests that old power structures—provided they are smart and willing to survive—will not be able to use it to their benefit

Jarvis 1.0 was all about celebrating Google, but Jarvis 2.0 has new friends in Facebook and Twitter. (An Internet intellectual always keeps up.) Jarvis 1.0 wrote that “Google’s moral of universal empowerment is the sometimes-forgotten ideal of democracy,” and argued that the company “provides the infrastructure for a culture of choice,” while its “algorithms and its business model work because Google trusts us.” Jarvis 2.0 claims that “by sharing publicly, we people challenge Google’s machines and reclaim our authority on the internet from algorithms.”

Jarvis has another reference point, another sacred telos: the equally grand and equally inexorable march of the Internet, which in his view is a technology that generates its own norms, its own laws, its own people. (He likes to speak of “us, people of the Net.”) For the Technology Man, the Internet is the glue that holds our globalized world together and the divine numen that fills it with meaning. If you thought that ethnocentrism was bad, brace yourself for Internet-centrism

Why worry about the growing dominance of such digitalism? The reason should be obvious. As Internet-driven explanations crowd out everything else, our entire vocabulary is being re-defined. Collaboration is re-interpreted through the prism of Wikipedia; communication, through the prism of social networking; democratic participation, through the prism of crowd-sourcing; cosmopolitanism, through the prism of reading the blogs of exotic “others”; political upheaval, through the prism of the so-called Twitter revolutions. Even the persecution of dissidents is now seen as an extension of online censorship (rather than the other way around). A recent headline on the blog of the Harvard-based Herdictproject—it tracks Internet censorship worldwide—announces that, in Mexico and Morocco, “Online Censorship Goes Offline.” Were activists and dissidents never harassed before Twitter and Facebook?

Most Internet intellectuals simply choose a random point in the distant past—the honor almost invariably goes to the invention of the printing press—and proceed to draw a straight line from Gutenberg to Zuckerberg, as if the Counter-Reformation, the Thirty Years’ War, the Reign of Terror, two world wars—and everything else—never happened.

even their iPad is of interest to them only as a “platform”—another buzzword of the incurious—and not as an artifact that is assembled in dubious conditions somewhere in East Asian workshops so as to produce cultic devotion in its more fortunate owners. This lack of elementary intellectual curiosity is the defining feature of the Internet intellectual. History, after all, is about details, but no Internet intellectual wants to be accused of thinking small. And so they think big—sloppily, ignorantly, pretentiously, and without the slightest appreciation of the difference between critical thought and market propaganda.

a possible structure recommended by one of the faculty we were talking to was – read/watch, quiz, lecture, student group discussion, reflection. The reasoning here is that if you give learners (particularly new learners) a reading without some form of accountability (a quiz) they are much less likely to do it. I know that for me, when I’ve done the readings, I’m far more likely to attend class. Putting the student group discussion after the lecture gives students who can’t attend a synchronous session a chance to review the recording

The standardization police have been telling us for years that each student must learn the same things. Poppycock. Scaffolding doesn’t mean taking away student choice. There are numerous approaches to allowing a little or a lot of choice into your classes (learner contracts come to mind). Just remember, most students don’t want choice – at first. 12-16 years of training has told them that you the faculty member have something you want them to do and they need to find the trick of it. It will take a while until those students actually believe you want their actual opinion.

You can have a goal like – get them acculturated to the field – and work through your activities to get there. It’s harder, they will need your patience, but once they get their minds around it, it makes things much more interesting.

Their training data is biased. They are designed by an unrepresentative group. They face the mathematical impossibility of treating all categories equally. They must somehow trade accuracy for fairness. And their biases are hiding behind millions of inscrutable numerical parameters.

biased AI systems can still be created unintentionally and easily. It’s also clear that the bias in these systems can be harmful, hard to detect and even harder to eliminate.

with North American computer science doctoral programs graduating only about 23% female, and 3% Black and Latino students, there will continue to be many rooms and many algorithms in which underrepresented groups are not represented at all.